[Paper Review] Conceptual Association for Compound Noun Analysis

Mark Lauer|ArXiv.org|Sep 6, 1994

Natural Language Processing Techniques4 citations

TL;DR

This paper proposes a corpus-based method for automatic syntactic disambiguation of compound nouns using conceptual associations derived from thesaurus categories. By extracting unambiguous compound instances from a corpus and leveraging statistical associations, the approach achieves 75% accuracy on a test set of 244 examples, with future work aiming to extend it to semantic role labeling for full interpretation.

ABSTRACT

This paper describes research toward the automatic interpretation of compound nouns using corpus statistics. An initial study aimed at syntactic disambiguation is presented. The approach presented bases associations upon thesaurus categories. Association data is gathered from unambiguous cases extracted from a corpus and is then applied to the analysis of ambiguous compound nouns. While the work presented is still in progress, a first attempt to syntactically analyse a test set of 244 examples shows 75% correctness. Future work is aimed at improving this accuracy and extending the technique to assign semantic role information, thus producing a complete interpretation.

Motivation & Objective

• To develop an automatic method for interpreting compound nouns in natural language processing.
• To address the challenge of syntactic ambiguity in compound nouns using corpus-derived statistical associations.
• To explore the use of thesaurus categories as a basis for conceptual association in disambiguation.
• To improve the accuracy of compound noun interpretation beyond syntactic structure alone.
• To lay the groundwork for extending the method to include semantic role information in future work.

Proposed method

• The method extracts unambiguous compound noun instances from a corpus to serve as training data for association patterns.
• Conceptual associations are built using thesaurus categories to represent semantic relationships between noun components.
• Statistical associations between head and modifier terms are computed based on co-occurrence frequencies in the corpus.
• The system applies these associations to resolve syntactic ambiguity in compound nouns by selecting the most probable interpretation.
• The approach uses a probabilistic ranking of interpretations based on accumulated association scores from the corpus.
• The method is evaluated on a test set of 244 compound nouns, with results reported as accuracy in syntactic disambiguation.

Experimental results

Research questions

• RQ1 How can corpus statistics be effectively used to disambiguate syntactic structures in compound nouns?
• RQ2 To what extent can thesaurus-based conceptual associations improve interpretation accuracy?
• RQ3 Can a statistical model based on unambiguous examples generalize to ambiguous compound noun forms?
• RQ4 What is the baseline performance of a corpus-driven, association-based method on a representative test set?
• RQ5 How might this method be extended to include semantic role labeling for full interpretation?

Key findings

• The method achieved a syntactic disambiguation accuracy of 75% on a test set of 244 compound noun examples.
• The use of thesaurus categories as conceptual anchors enabled effective statistical association learning from corpus data.
• Unambiguous compound noun instances were successfully extracted and used to build a reliable association model.
• The approach demonstrated feasibility for automatic compound noun interpretation using only corpus statistics and lexical categorization.
• The results suggest that conceptual associations can serve as a strong foundation for further extension to semantic role assignment.
• The method remains limited to syntactic disambiguation in this version, with semantic interpretation as a planned future extension.